boxer3d

An iPhone app that detects objects in AR and places 3D bounding boxes around them, using LiDAR, YOLO and BoxerNet. A demo of 3D object detection on a phone.

Share on XNo license declared

Overview

Boxer3D is an iPhone app that detects everyday objects and draws 3D oriented bounding boxes around them in augmented reality. It uses YOLO11n for 2D detection, then lifts the results to 3D with Meta Research's BoxerNet, using LiDAR depth and ARKit camera pose. The boxes are rendered as wireframes with SceneKit, anchored in the real world. It works as a demo of 3D detection running on a phone.

Key features

  • YOLO11n detects objects across 80 COCO classes
  • BoxerNet lifts 2D boxes to 3D oriented boxes
  • Uses LiDAR depth and ARKit pose and gravity
  • SceneKit draws wireframe boxes in AR

Best for

Developers exploring on-device 3D object detection on iOS. It needs an iPhone 12 Pro or later with LiDAR, iOS 16 or newer, and about 450 MB for models.

Upstream
Barath19/Boxer3D
Fork on GitHub
Guo-astro/boxer3d
Upstream stars
429
Category
Finance, data and research
Language
Swift
License
No license declaredWithout a license, the author keeps all rights. Ask the upstream owner before reusing the code.
Forked
2026-04-13
Sync status
In syncLast synced 2026-09-29